A data center consumes a large amount of electricity, and currently, a power level of electricity use of a data center ranges from several megawatts to several hundreds of megawatts. As the information technology (IT) industry develops, a power of electricity use of a data center increases continuously. Therefore, electricity expenses occupy a considerably large proportion in operation costs of a data center. It is of great significance to reduce electricity expenses of a data center.
In other approaches, there is a method for controlling electricity expenses of a data center, and the method is applied to a time-of-use price. In a high-electricity-price period (that is, a peak period of electricity consumption) of each cycle (usually one day), load is decreased, and in a low-electricity-price period (that is, a trough period of electricity consumption), load is increased, thereby reducing electricity expenses of a data center.
As to a time-of-use price, generally, the government specifies the peak period (for example, 8 o'clock to 22 o'clock) of electricity consumption and the trough period (for example, 22 o'clock to 8 o'clock of the next day) of electricity consumption, and determines an electricity price of each period. The peak period of electricity consumption and the trough period of electricity consumption are obtained according to empirical statistics collection in the long run, and are relatively constant and will not be changed arbitrarily.
In the other approaches, a data center adjusts load based on the peak period of electricity consumption and the trough period of electricity consumption. However, the peak period of electricity consumption and the trough period of electricity consumption are relatively constant and both last for a relatively long time, but a quantity of services changes in real time, and load also changes in real time. Consequently, a cycle of adjusting load by a data center is relatively long. This affects load adjustment precision, and does not help control costs of electricity expenses.